Introduction

Heartburn is a pain in the chest felt when acid from the stomach travels up the oesophagus. This can happen as a result of different factors including smoking, being overweight, consumption of certain food, alcohol consumption, and stress [1]. To aliviate heartburn, Gaviscon can be administered. Using data on monthly prescribed medication, this study will investigate the change in Gaviscon prescription during the Winter holidays. December, with the festivities, is a month well associated with the over consumption of food and alcohol, along with increased stress [2]. After the New Year celebrations, however, this changes [2]. This might suggest that Gaviscon use would be higher during December compared to January when people return to normal eating, drinking, and stress levels, or take on new health regime. As well as observing the general trends in prescription of Gaviscon, there will also be a focus on how COVID lockdown or the education level of a population might have affected prescriptions during the Winter holidays.

Loading Libraries

library(tidyverse)
library(janitor) #clean data
library(here) #locate data
library(gt) #tables
library(grafify) #colour scales for plot
library(plotly) #for interactive plotting
library(forcats) #for reordering values

Loading data

Importing data files of December and January 2019-2023 downloaded from https://www.opendata.nhs.scot/dataset/prescriptions-in-the-community . These years were chosen to compare winter holidays with and without COVID lockdown which restricted social events, changing the dynamic of the usual celebrations [3]. Health board data, including health board names, can be downloaded from https://opendata.scot/datasets/public+health+scotland-geography+codes+and+labels/. The Scottish census, done in 2022, provides information on the qualification levels of the population. Data on the levels of education can be found at https://www.scotlandscensus.gov.uk/webapi/jsf/tableView/tableView.xhtml.

#creating a list of the files containing prescription data
files <- list.files(here("data"), pattern = "20") 
#pattern of 20 will use only the data with the number 20 in it, which is only the prescription data in this case

#creating a dataframe which combines all of the information in the files and cleaning the column names for consistency
all_data <- files %>% 
  map_dfr(~read_csv(here("data", .))) %>%
  clean_names() %>% 
  mutate(hb = coalesce(hbt, hbt2014)) 
#joining the information from two differently named healthboard code columns

#loading in data for healthboard names and cleaning the column names for consistency
hb <- read_csv(here("data", "hb14_hb19.csv")) %>% 
  clean_names()

#joining precription data to healthboard names
hb_and_prescription <- all_data %>% 
  full_join(hb, by = c("hb" = "hb"))

#loading in data for education level of healthboard
education_levels <- read_csv(here("data", "table_2024-11-22_21-18-31.csv"), skip = 10) %>% 
  #skipping the first 10 rows as they contain irrelevant information
  clean_names() %>% 
  filter(row_number() <= n()- 3) 
#removing the last 3 rows as they contain irrelevant information
#a warning message shows, the problematic rows do not contain information needed and are removed

Wrangling data for the initial plot

#The data must be wrangled in order to make it easier to use later.
#formatting the dates for easy use later on
combined_scotland_data  <- hb_and_prescription %>% 
  mutate(paid_date_month = parse_date_time(paid_date_month, "ym")) %>% 
  mutate(year = format(paid_date_month, "%Y"))

#creating a function for the creation of date columns needed, as this is required multiple times
dates_func <- function(df) {
  df %>% 
  mutate(holiday = case_when(month(paid_date_month) == 12 ~ paste(year(paid_date_month), year(paid_date_month) + 1, sep = "/"),
                             month(paid_date_month) == 1 ~ paste(year(paid_date_month) - 1, year(paid_date_month), sep = "/"))) %>% 
  mutate(month = month(paid_date_month, label = TRUE, abbr = FALSE))
}

#preparing prescription data for plotting
for_plot <- combined_scotland_data %>% 
  filter(str_detect(bnf_item_description, "GAVISCON")) %>% 
  group_by(paid_date_month) %>% 
  summarise(paid_quantity = sum(paid_quantity)) %>% 
  dates_func() %>% 
  mutate(month = fct_reorder(month, desc(paid_quantity))) 
#reordering the months by paid quantity values to allow bar chart to ensure longer bars are behind shorter bars when plotting

Plotting prescription data

#plotting prescription data
gaviscon <- for_plot %>% 
  ggplot(aes(x = holiday, y = paid_quantity, fill = month))+
  geom_bar(stat = "identity", position = "identity", alpha = 0.7)+ 
  #position is specified to ensure that the month values are not stacked upon each other
  scale_fill_manual(values = rep(c("purple", "#FFD700")))+
  theme_minimal()+
  labs(x = "Winter holiday during which prescribed", 
       y = "Quantity of Gaviscon Prescribed", 
       title = "Trend in Gaviscon prescription over winter holidays in Scotland",
       fill = "Month")

#making the prescription data interactive
interactive <- gaviscon %>% 
  ggplotly()
interactive

From this plot we can see that each observed year shows a decrease in the amount of prescription of Gaviscon from December into January in Scotland. This supports the idea that over-consumption during December leads to higher prescription of Gavsicon compared to January. It also points to lockdown having an effect on the health of the population during the 2020-2021 winter holidays compared to other years - December prescriptions being higher as well as the difference from December to January.

Comparison of differences, across Scotland, between December and January

Here, the numerical differences in Gaviscon prescribed in December and the following January are calculated and can be compared against each other as a percentage change.

#preparing data for making a table that compares December and January paid quantity
change_in_prescription <- combined_scotland_data %>% 
  filter(str_detect(bnf_item_description, "GAVISCON")) %>% 
  group_by(paid_date_month) %>% 
  summarise(paid_quantity = sum(paid_quantity)) %>% 
  dates_func() %>% 
  select(holiday, paid_quantity, month) %>% 
  pivot_wider(names_from = month, values_from = paid_quantity) %>% 
  mutate(difference = January - December) %>% 
  mutate(percent = difference/December)

#creating table
change_table <- change_in_prescription %>% 
  gt() %>% 
  fmt_number(columns = c(December, January, difference), decimals = 0) %>% 
  fmt_percent(columns = percent, decimals = 2) %>% 
  tab_header(title = "Prescription Quantity of Gaviscon During Winter Holidays", 
             subtitle = "Data from NHS Scotland") %>% 
  cols_label(holiday = " ",
             difference = "Change", 
             percent = "Percentage Change") %>% 
  cols_align(align = "center",
              columns = percent)
change_table
Prescription Quantity of Gaviscon During Winter Holidays
Data from NHS Scotland
December January Change Percentage Change
2019/2020 9,171,039 8,711,055 −459,984 −5.02%
2020/2021 9,811,769 8,390,883 −1,420,886 −14.48%
2021/2022 8,713,855 7,382,644 −1,331,211 −15.28%
2022/2023 8,027,304 7,267,184 −760,120 −9.47%

This table confirms that the change in Gaviscon prescription increased during the 2020-2021 winter holidays. This suggests that lockdown had an effect on the health habits of the Scottish population. However, a larger increase is observed the following year before it drops again. The 2021/2022 holidays still involved some restrictions due to the pandemic, therefore it is still possible that the change in usual celebrations had an effect on Gaviscon prescription.

Investigating the differences between healthboards

Education level has been proven to have an effect on the quality of eating and drinking habits [4]. Therefore, it may be possible that the level of education will influence the change in Gaviscon prescription. A greater knowledge of health may result in a healthier lifestyle in January, leading to less prescription of Gaviscon. However, it may also result in more awareness of the change in lifestyle during December and show a similar level of prescription between the two months. The effect of education on the change in prescription between December and January of Gaviscon will be investigated.

#preparing data to show the percentage of people in each health board without qualifications
cleaning_edu <- education_levels %>% 
  slice(1:15) %>% 
  filter(!is.na(highest_level_of_qualification)) %>% 
  filter(!is.na(all_people_aged_16_and_over)) %>% 
  mutate(all_people = as.numeric(all_people_aged_16_and_over)) %>% 
  mutate(proportion = (no_qualifications/all_people)*100) %>% 
  #calculating the percentage of people without qualifications
  mutate(hb_name = paste("NHS", highest_level_of_qualification))

#finding the average difference in paid quantity between December and January by health board
gaviscon_by_hb <- combined_scotland_data %>% 
  filter(str_detect(bnf_item_description, "GAVISCON")) %>%
  dates_func() %>% 
  group_by(hb_name, month) %>% 
  summarise(paid_quantity = sum(paid_quantity)) %>% 
  pivot_wider(names_from = month, values_from = paid_quantity) %>% 
  mutate(average_diff = (December - January)/4) 
#dividing the total differences by 4 to get an average over the four years observed

#joining data and scaling paid quantity to the population of each health board
gaviscon_and_education <- gaviscon_by_hb %>% 
  full_join(cleaning_edu) %>% 
  mutate(gaviscon_proportion = average_diff/all_people)

#plotting the percentage of unqualified people against the average change in Gaviscon quantity prescribed by health board
hb_gav_plot <- gaviscon_and_education %>% 
  ggplot(aes(x = proportion, y = gaviscon_proportion, colour = hb_name))+
  geom_point(size = 3)+
  theme_minimal()+
  scale_colour_grafify(palette = "kelly")+
  labs(title = "Investigating correlation between education levels and decreased\n Gaviscon prescription", 
       y = "Average decrease in prescription",
       x = "Percentage of individuals with no qualifications", 
       colour = "Health Board")
hb_gav_plot

This plot shows no obvious correlation between the level of education and the change of Gaviscon prescription over the winter holidays.

Conclusion

Overall, this exploration of the change in Gaviscon prescription, in Scotland, has suggested that there may be an effect between different months of the winter holidays. There appeared to be a slightly larger decrease in Gaviscon prescription during January compared to that of all prescriptions. The difference in prescription observed in 2020/2021, during lockdown, was increased from the previous year, suggesting that it is not the social aspect of December, but possibly the stress that leads to higher Gaviscon prescription. There was no clear association between qualification level and change in Gaviscon prescription. This assessment only accounted for people with no qualifications, however involving other low levels of qualification may indicate a correlation. This investigation is limited by the amount of data used, a further exploration into a larger number of years would need to be done to have more confidence in the trends observed. Statistical analysis would also need to be done for a certain conclusion to be made, this could involve obtaining effect sizes and their standard deviations.

Use of Generative AI

Generative AI was used to identify prescribed medications that would show trends in prescription throughout the year. This was used to inspire the current investigated topic.

References